Research On Acoustic Signal Detection And Recognition Technology For Oil And Gas Pipeline Leakage

While the number of natural gas pipelines has increased and the operating years have increased,natural gas pipeline leakage occurs from time to time and can easily cause serious accidents.Therefore,it is very important to monitor the running status of the pipeline in real time.In this paper,based on the acquisition,denoising,feature extraction and gas pipeline identification as the research object,through the experimental analysis,the frequency range of the leakage acoustic signal of the gas pipeline in various states is determined.The wavelet threshold denoising algorithm is studied and the threshold function is improved,the VMD algorithm is studied,the VMD-Wavelet denoising algorithm is proposed,the various characteristic parameters of the acoustic signal of the gas pipeline are studied,and the method of extracting the feature entropy of cloud model based on VMD-Wavelet is proposed.The BP neural network model is studied and the VMD-En-BP neural network model is proposed.The main work of this article is as follows:Firstly,the basic principle of acoustic leakage detection is studied,and the leak detection system of gas pipeline is designed.The time-domain and frequency-domain characteristics of leakage signals in different states are analyzed,and it is concluded that the energy of leakage acoustic signals is mainly concentrated in low frequency band.Secondly,studied the soft threshold and hard threshold and Garrote threshold and other classical wavelet thresholding function,an improved wavelet threshold function is proposed.By comparing the improved wavelet threshold function with the classical wavelet threshold function,the results show that the improved wavelet threshold function can achieve better denoising effect in SNR and MSE.Thirdly,the VMD algorithm,Hausdorff distance,kurtosis and wavelet transform are studied,and the VMD-Wavelet denoising method is proposed.Firstly,the noisy signal is decomposed by VMD,and the effective modal component is selected by the Hausdorff distance.Calculate the kurtosis value of the unselected modal component and filter out the component whose kurtosis is greater than the selected threshold.The high frequency noise of the component is filtered out by the improved wavelet threshold function in this paper.Finally,the processed IMF components are used to reconstruct the signal.The theoretical analysis and simulation results show that this method not only has a good denoising effect on typical signals,but also has a good denoising effect on the leakage signal,tapping signal and normal signal of actual gas pipeline.Then,through experiments,this paper verified that EMD decomposition could not accurately extract features from the spectrum of modal components of acoustic signals in various states,while IMF1 and IMF2 after VMD decomposition could reflect the differences in the center frequencies of acoustic signals in various states.Therefore,a cloud model feature entropy extraction method based on VMD-Wavelet is proposed in this paper.The method uses VMD-Wavelet algorithm to denoising the acoustic signals collected under the three states of normal operation,percussion and leakage of the gas pipeline.K effective modal reconstructed signals are selected to calculate the cloud model characteristic entropy of reconstructed signals.Simulation analysis shows that the cloud model characteristic entropy of reconstructed signals in three states calculated by this method has a good degree of discrimination and can be used as the characteristic parameter of state recognition.Finally,BP neural network is studied and the VMD-BP and En-BP neural network models are constructed.According to their advantages and disadvantages,the VMD-En-BP neural network model is put forward,which is used to identify the running state of the gas pipeline.The experimental results show that the model can accurately identify the normal operation,percussion,leakage,small leakage and large leakage of the gas pipeline.